Adoption of a New Payment Method:
Theory and Experimental Evidence
Jasmina Arifovicy
John Duffyz
Simon Fraser University
University of California, Irvine
Janet Hua Jiangx
Bank of Canada
September 30, 2015
Abstract
We model the introduction of a new payment method, e.g., e-money, that competes
with an existing payment method, e.g., cash. The new payment method involves relatively lower per-transaction costs for both buyers and sellers but sellers must pay a
one-time fixed fee to accept the new payment method. Due to network effects, our
model admits two symmetric pure strategy Nash equilibria. In one equilibrium, the new
payment method is not adopted and all transactions continue to be carried out using the
existing payment method. In the other equilibrium, the new payment method is adopted
and completely replaces the existing payment method. The equilibrium involving only
the new payment method is socially optimal as it minimizes total transaction costs. Using this model, we study the question of equilibrium selection by conducting a laboratory
experiment. We find that, depending on the fixed fee charged for adoption of the new
payment method and on the choices made by participants on both sides of the market,
either equilibrium can be selected. More precisely, a lower fixed fee for sellers favors
very quick adoption of the new payment method by all participants while for a sufficiently high fee, sellers gradually learn to refuse to accept the new payment method and
transactions are largely conducted using the existing payment method.
JEL classification numbers: E41, C35, C83, C92
Keywords: Payment methods, network effects, experimental economics.
For their comments and discussions, we would like to thank Gerald Stuber and participants at the 2015 CEA
meetings, the 2015 Barcelona GSE Summer Forum on Theoretical and Experimental Macroeconomics, the 1st
International Conference of Experimental Economics in Osaka, the 2015 Econometric Society World Congress,
and the Bank of Canada. The views expressed in this paper are those of the authors. No responsibility for them
should be attributed to the Bank of Canada.
y
[email protected]
z
[email protected]
x
[email protected]
1
Introduction
The payments industry has undergone significant changes in the past few decades. Various new means of payments that compete with the traditional payment method – cash – have
entered into the payment landscape, including debit cards, credit cards, general purpose prepaid cards such as Visa and MasterCard gift cards, public transportation cards that expand
into retail transactions such as the Octopus card in Hong Kong, mobile payments such as MPesa in Kenya, online money transfer schemes such as Paypal, and virtual crypto-currencies
such as the Bitcoin. The numerous and varied attempts to introduce new payment methods
have met with mixed results. The Octopus card in Hong Kong is a notable success. Danmont
in Denmark ultimately failed after a few years of limited success. Mondex debuted with
much fanfare in several countries but soon failed.
In this paper we study the introduction of a new payment method, focusing on electronic
pre-paid schemes. We seek answers to several questions. Will a socially more efficient new
payment method take off? Will merchants accept it? How do consumers make portfolio
choices between the existing and the new payment method? How do merchants and consumers interact with each other? How do the cost structure associated with the new payment
system influence the result?
We first develop a model of the introduction of a new payment instrument, "e-money,"
that competes with an existing payment method, "cash." We model the new payment method
as being more efficient for both buyers and sellers in terms of per transaction costs. Such
cost-saving motive lies behind the various attempts to introduce a new payment method:
after all, if the new payment method did not offer any such cost savings, there would be
no reason to expect it to be adopted (or even introduced). There is also evidence that new
developments in electronic payment technology can greatly speed up transactions and save on
various handling costs associated with traditional payment methods. According to a study by
Polasik et al. (2013), who analyze the speed of various payment methods from video material
recorded in the biggest convenience store chain in Poland, a transaction using contactless
cards in offline mode without slips costs on average 25.71 seconds, a significant reduction
over a cash transaction, which costs 33.34 seconds. Arango and Taylor (2008) suggest that,
after the various cash-handling costs – including deposit reconciliation, deposit preparation,
deposit trips to banks, coin ordering, theft and counterfeit risk, etc. – are accounted for, a
cash transaction costs merchants $0.25 and a (PIN) debit transaction costs $0.19.1 Working
against the adoption of the new e-money payment method, we assume that sellers have to pay
1
The calculation is based on a transaction value of $36.5. The cost of the debit transaction includes a $0.12
payment-processing fee.
1
a fixed setup fee (to rent or purchase a terminal) to process the new payment method.
Given the cost structure of the two payment methods, buyers and sellers play a two-stage
game. In the first stage, agents make payment decisions simultaneously. Buyers allocate
their budget between the new and existing payment methods. Sellers decide whether to pay
the fixed cost to accept the new payment method (they always accept the existing payment
method). The second stage consists of multiple rounds of meetings where buyers and sellers
trade with each other. In each meeting, the buyer observes the seller’s payment choice, and the
trade is successful if the buyer uses a payment method accepted by the seller. Due to network
effects, the model admits two symmetric pure strategy Nash equilibria. In one equilibrium,
the new payment method is not adopted and all transactions continue to be carried out using
the existing payment method. In the other equilibrium, the new payment method is adopted
and completely replaces the existing payment method. The equilibrium involving only the
new payment method is socially optimal as it minimizes total transaction costs (the reduction
in per transaction cost exceeds the fixed cost paid by sellers).
As our model has multiple equilibria, we next conduct a laboratory experiment to assess
conditions under which the new payment method replaces the existing payment method, and
also conditions where the new payment method fails to be adopted. We find that, depending
on the fixed cost for the adoption of the new payment method and on the choices made by
participants on both sides of the market, either equilibrium can be selected. More precisely,
a lower fixed cost for sellers favors very quick adoption of the new payment method by all
participants while for a sufficiently high fee, sellers gradually learn to refuse to accept the new
payment method and transactions are largely conducted using the existing payment method.
We view the experimental approach as a useful complement to theoretical and empirical
research on the acceptability of payment methods. The theoretical literature emphasizes the
importance of network externalities in the adoption of new payment methods (Rochet and
Tirole, 2002; Wright, 2003; McAndrews and Wang, 2012; Chiu and Wong, 2014). For the
consumer (merchant), the benefit of adopting (accepting) a new payment method increases
if more consumers use (merchants accept) the payment method. These network effects lead
to multiple equilibria, which poses a problem for theoretical predictions and thus our experimental study can shed some light on which equilibrium is more likely to occur. While
the theory often focuses on equilibrium analysis and ignores transition dynamics, our experimental approach also provides some useful insight about the process by which a new payment
method may take off and the speed with which this may occur.
There is a large empirical literature using survey data to explore individual choices among
2
different means of payments.2 These studies focus mainly on choices among existing payment methods, and the decisions made by a single party, either the consumer or the merchant.
In our model and experiment, we develop an environment suitable for the study of new payment methods that considers interactions between both sides of the market (buyers and sellers). In addition, survey data are subject to errors due to insufficient incentives for truthful
or careful reporting, or misunderstandings about the survey questions; these problems are
alleviated to some extent in an incentivized experimental study.3 Finally, since many factors
are at play in the field, isolating the effect of a particular factor may be challenging. In the
laboratory, we can exert better control over the environment and thereby isolate the factors
that play a role in whether or not a new payment method is adopted.
The closest paper to this one is by Camera, Casari and Bortolotti (2015, hereafter CCB),
who also develop a model of payment choice between cash and e-money/card and conduct
an experimental study. While we take inspiration from CCB’s paper, our project differs
from theirs along several important dimensions, including theoretical construct, experimental designs, research questions and experimental results. CCB study how the presence of
proportional seller fees and buyer rewards affects the adoption of card payments, assumed
to be more "reliable" than cash.4 They find that sellers readily adopt cards disregarding the
fee and reward structure, while buyers are sensitive to these incentive schemes. More specifically, the buyer’s adoption rate is high in the absence of fees or rewards. Imposing seller fees
alone reduces card adoption, but adding buyer rewards neutralizes the effect of seller fees
and restores high card adoption. CCB also find that there is little feedback effect between the
two sides of the market and that the network effect plays no role in their experiment. Our
research question is how the adoption of a new and more socially efficient payment method
is affected by the fixed cost born by sellers relative to the potential saving on per transaction
costs. We find that the choices by both buyers and sellers depend critically on the magnitude
of the fixed cost. At the same time, we observe a strong feedback effect between the two
2
See, for example, Arango et al. (2011) for analysis with the Bank of Canada 2009 Method of Payment
Survey, Schuh and Stavins (2013) with the Federal Reserve Bank of Boston 2008 U.S. Consumer Payment
Choice Survey, and von Kalckreuth et al. (2014) with the Deutsche Bundesbank 2008 Payment Habits in
Germany Survey. Recently, Bagnall et al. (2015) study consumers’ use of cash by harmonizing payment diary
surveys from seven countries.
3
Klee (2008) and Cohen and Rysman (2012) study payment choices in grocery stores using scanner data.
Misreporting does not pose a problem for scanner data, but it shares other features of the survey data.
4
CCB model e-money as more efficient in terms of per transaction cost, which is similar to us but in a
different sense. They assume that cash payments are "manual" and consequently less reliable than electronic
payments. To implement this in the laboratory, they require that buyers using cash must manually click on the
correct combination of bills of different denominations within a set time limit, while a card payment is quickly
done with a single mouse click. This is somewhat unrealistic. We rarely see stores turning away customers due
to slow payment processing. Stores usually deal with the problem by hiring more cashiers, which we capture
by a higher per transaction cost.
3
sides.
There are several possible reasons why subjects react to the other side’s choices so strongly
in our experiment, but not in CCB. First, the existence of the fixed setup cost in our model
makes sellers respond more strongly to buyers’ cash holdings. In particular, they would reject e-money if buyers hold only cash. In CCB, although the homogeneous adoption of cash
payments consists of a Nash equilibrium, it is not robust to trembles: sellers are indifferent
between acceptance and rejection in this equilibrium and have the incentive to accept cards to
minimize payment mismatch if there is no cost to accept cards. This could explain why seller
adoption is consistently high in CCB disregarding buyers’ choices. Second, in CCB, subjects
make both payment and terms of trade decisions. Such multi-tasking may be too challenging
for subjects and cause them to focus on the more salient fee/reward structure rather than the
coordination problem.5 For example, consider the treatment with fees and no rewards. Relative to the all-cash equilibrium, the all-card equilibrium is characterized by a higher trading
probability (because card transactions are more reliable), a higher price (because sellers pass
fees onto buyers) and a lower quantity per trade. In terms of payoffs, buyers are better off
in the all-card equilibrium despite the higher price, while sellers are worse off.6 In theory,
buyers should push for card adoption while sellers should resist it. However, the opposite
is observed in the experiment, where the buyer’s adoption rate is significantly lower. This
suggests that the buyers are predominantly concerned that sellers will pass fees to them and
fail to take advantage of the high seller acceptance rate. While we agree that the setting of
the terms of trade is an interesting question, we think that it is better studied separately. In
our experiment, we fix the terms of trade so that subjects can concentrate on the choice of
payment methods and the related coordination problem. Third, in our experiment, buyers
meet all sellers and vice versa in each period and subjects quickly learn about the aggregate
market condition, which boosts interaction between the two sides. In CCB, subjects visit a
different partner in every period and observe only their own trading histories. The lack of
market-level information may slow down interaction across the two sides.
The rest of the paper is organized as follows. Section 2 develops the theoretical model.
Section 3 describes our experimental design. The experimental results are presented in Section 4. Finally, section 5 concludes and points out some directions for future research.
5
The observation that the price and quantity realized in the experiment are far away from theoretical predictions suggests that subjects have difficulty in fully optimizing during the experiment.
6
CCB leave it to subjects to figure out their expected payoffs conditional on other players’ choices, which
may be a daunting task.
4
2
The Model
In this section, we develop a simple model of the adoption of a new payment method,
“e-money,” that competes with an existing payment method, “cash.” In each trading period,
buyers make a portfolio decision, splitting their budget between e-money and cash. Simultaneously, sellers decide whether or not to accept e-money transactions; cash payments must
always be accepted. Buyers then meet with sellers and engage in transactions using one payment form or the other. For simplicity (and later experimental implementation) we assume
homogeneous buyers and sellers, an exogenously given spending budget for the buyer and
fixed terms of trade. As in other models of payment competition, our model has multiple
equilibria.
We model the new payment method, e-money, as being more efficient for both buyers
and sellers in terms of per transaction costs, but sellers must pay a fixed cost to process emoney. This setup cost, however, can be recovered if each seller succeeds in conducting
a sufficient number of e-money transactions with buyers. Thus, both transaction costs and
network effects play a role in decisions to adopt the new payment method. In our model, the
adoption of an e-money payment method is the socially efficient outcome. We now turn to a
detailed description of our model.
2.1
Physical Environment
There are a large number of buyers and sellers (or stores) in the market, each of unit
measure. Each seller i 2 [0; 1] is endowed with 1 unit of good i. The seller derives zero
utility from consuming his own good and instead tries to sell his good to buyers. The price
of the good is fixed at one. Each buyer j 2 [0; 1] is endowed with 1 unit of money. In
each period, the buyer visits all sellers in a random order. The buyer would like to consume
one and only one unit of good from each seller, and the utility from consuming each good is
u > 1.
There are two payment instruments: cash and e-money; we sometimes refer to the latter
as “cards.” Each cash transaction incurs a cost, b ; to buyers, and a cost, s ; to sellers. The per
transaction costs for e-money are eb and es for buyers and sellers, respectively. Sellers have
to pay an up-front cost, F > 0; that enables them to accept e-money payments, for example,
to rent or purchase a terminal to process e-money transactions.
In the beginning of each trading period, sellers decide whether to accept e-money at the
one-time fixed cost of F , or not. Cash, being the traditional (and legally recognized) payment
5
method, is universally accepted by all sellers. Simultaneous with the sellers’ decision, buyers
make a portfolio choice as to how to divide their money holdings between cash and e-money.
After sellers have made acceptance decisions and buyers have made portfolio decisions, the
buyers then go shopping, visiting all of the stores in a random order. When a buyer enters
store i, he buys one unit of good i if the means of payment are compatible. Otherwise, there is
no trade. At the end of the trading period, both buyers and sellers spend their money balances
on a general good. One unit of the general good costs one dollar and entails one unit of utility.
Buyers do not wish to consume the general good and any unspent money does not yield them
any extra utility.7
In what follows, we make the following four assumptions about costs:
A1:
e
b
< b and
and sellers.
A2: u
c
b
e
s
<
s.
In words, e-money saves on per transaction costs for both buyers
> 0. Under this assumption, buyers prefer cash trading to no trading.
e
e
A3: F
s
s + b
b . This condition implies that the net benefit of investing in the
ability to process e-money transactions is positive for the society if all transactions are
carried out in e-money.
A4: F
1
is used.
2.2
e
s.
This assumption ensures the existence of an equilibrium where e-money
Equilibrium
We will focus on symmetric equilibria, where all buyers make the same portfolio choice
decision, and all sellers make the same e-money acceptance decisions. Let 0 mb 1 be
the e-money balance chosen by the buyer, and let 0 ms 1 denote the fraction of sellers
who accept e-money. If 0 < ms < 1, then sellers play a mixed strategy, accepting e-money
with probability ms .
2.2.1
Buyer’s Decision
We will first analyze the buyer’s decision, mb , conditional on the seller’s adoption decision ms . We will carry out the analysis in two cases: (1) mb ms , and (2) mb ms .
7
The assumptions that sellers do not value their own goods and buyers do not value the general good are for
simplicity. The model’s implications do not hinge on these assumptions.
6
If mb ms , then each buyer makes ms purchases using e-money and 1 mb purchases
using cash. Buyers are not able to transact with a fraction mb ms of sellers due to payment
mismatches (buyers want to use e-money but sellers accept cash only). The buyer’s expected
payoff in this case is:
e
) + (1 mb )(u
b = ms (u
b)
| {z b} |
{z
}
e-money transaction
cash transaction
Note that
d b =dmb =
(u
b)
< 0:
It follow that for this case, the optimal choice of each buyer is to reduce mb to ms so as to
minimize the probability of a payment mismatch or no trade outcome.
If mb ms , then each buyer makes mb e-money transactions and m mb cash transactions (among which ms mb are with sellers who also accept e-money). The buyer’s expected
payoff is now given by:
b
= mb (u
| {z
e
b)
}
e-money transaction
In this case we have that
+ (1
|
e
b
d b =dmb =
mb )(u
{z
cash transaction
+
b
b) :
}
> 0:
Thus, if mb ms ; then buyers should increase their e-money balances to ms so as to minimize transaction costs.
From the analysis above it follows that buyers’ optimal portfolio decision is to mimic the
sellers’ acceptance decision:
mb (ms ) = ms :
2.2.2
Seller’s Decision
We now turn to the seller’s acceptance decision conditional on the buyer’s portfolio decie
sion, mb . We will carry out our analysis under two parameter settings: (1) F
s
s , and
e
(2) F
s
s . For each parameter setting, similar to the discussion of the buyer’s choice,
we analyze the seller’s decision in two cases: mb ms and mb ms .
e
Parameter Setting I: F
If mb
ms , then each seller who accepts e-money
s
s
engages in a unit measure of e-money transactions (remember that buyers use e-money when-
7
ever the seller accepts it), and has a payoff of
e
s
e
s
=1
F:
Sellers who only accept cash engage in an average of (1 mb )=(1 ms ) 1 cash transactions
(the total cash balance in the economy is 1 mb and this is divided among the 1 ms sellers
who only accept cash). The sellers who only accept cash thus have a payoff of
s
=
1
1
mb
(1
ms
s ).
In this case,
e
s
!
1 mb
e
1
F =
(1
s) !
s
1 ms
(1 mb )(1
s)
ms (mb ) = 1
e
1
F
s
1
e
[ s
F + mb (1
=
s
e
1
F
s
=
s
s )]
(1)
e
0
If F < s
s , then equation (1) has a positive intercept and lies above the 45 line. As long
as mb
ms , each seller who accepts e-money is able to use e-money during all meetings,
which makes it profitable to pay the fixed cost, F; to accept e-money. In equilibrium, it must
be the case that mb ms .
If mb ms , the e-money balance in the economy can support mb e-money transactions,
which are divided among ms sellers who accept e-money. Each seller who accepts e-money
can trade in all meetings, among which mb =ms will be e-money transactions, and the remaining 1 mb =ms will be cash transactions. The expected payoff of a seller who accepts
e-money is therefore:
e
s
=
mb
(1
m
| s {z
e
s)
+ 1
}
|
e-money transaction
= (1
mb
(
s) +
ms
s
mb
(1
ms
{z
cash transaction
e
s)
F:
s)
F
}
Sellers who only accept cash engage in cash transactions in all meetings and have a payoff of
s
=1
8
s.
In this case,
e
s
=
s
s
ms (mb ) =
!
e
s
F
mb .
(2)
e
e
If F < s
F=( s
s , then equation (2) has a slope that is > 1. If mb
s ), then it is
e
a dominant strategy for sellers to accept e-money: each seller makes more than F=( s
s)
e
e-money sales to warrant the fixed investment for e-money acceptance. If mb F=( s
s ),
the number of e-money transactions is not large enough to recover the fixed acceptance cost
e
for all sellers. As a result, sellers play a mixed strategy: ms = mb ( s
s )=F fraction
of sellers accept e-money, and the rest accept cash only. All sellers earn the same expected
payoff ( s = es ).
To summarize, if F <
such that
e
s;
s
then given the buyer’s strategy mb , the seller’s strategy is
ms (mb ) =
8
<
mb (
e)
s
s
F
: 1
F
if mb
s
if mb
s
e
s
F
e
s
;
:
e
e
Parameter Setting II: F > s
If F > s
ms , equation
s
s ; in the case where mb
0
(1) has a negative intercept and lies below the 45 line. It is a dominant strategy for sellers
e
F ]=(1
m
^ b , then accepting
not to accept e-money if mb m
^ b 1 [(1
s ). If mb
s)
e-money gives a higher payoff iff ms ms (mb ); in equilibrium, sellers play a mixed strategy
choosing to accept with probability ms (mb ).
If mb ms , then the slope of equation (2) is < 1. This implies that as long as mb ms ,
sellers who do not accept e-money earn a higher payoff. As a result, ms will decrease. In
equilibrium, it must be the case that mb ms .
To summarize, under the parameter setting F > (
the seller’s strategy is such that
8
< 0
ms (mb ) =
:
(1
2.2.3
1
e)
s
F
[(
s
e
s)
F + mb (1
s
e
s );
s )]
given the buyer’s strategy mb ,
if mb
1
(1
if mb
1
(1
e)
s
s
F
1
;
e)
s
s
F
1
:
Equilibrium
Combining the analysis above, we can characterize the symmetric equilibrium of the
economy using Figure 1. There are at least two symmetric pure strategy equilibria. In one of
9
these equilibria, mb = ms = 1: all sellers accept e-money, and all buyers allocate all of their
endowment to e-money – call this the all-e-money equilibrium (this equilibrium always exists
e
provided that F
1
s ). There is a second symmetric pure strategy equilibrium where
mb = ms = 0 and e-money is not accepted by any seller or held by any buyer – call this the
all- cash equilibrium. In both equilibria, there is no payment mismatch, and the number of
e
transactions is maximized at 1. In the case where F = s
s ; there exists a continuum of
possible equilibria in which ms 2 (0; 1) and mb = ms :
The e-money equilibrium is socially optimal as it minimizes total transactions cost. Note
that buyers are always better off in the all- e-money equilibrium relative to the all cash equilibrium. The seller’s relative payoff in the two equilibria, however, depends on the fixed cost,
e
F , and on the savings on per transaction costs from the use of payment 2. If F = s
s,
e
then the seller’s payoff is the same in the cash and e-money equilibria; if F < s
s , then
the seller’s payoff is higher in the e-money equilibrium than in the cash equilibrium; finally,
e
if F > s
s , then the seller’s payoff is lower in the e-money equilibrium than in the cash
equilibrium.
10
3
Experimental Design
For each session of our experiment, we recruited 14 inexperienced subjects and randomly
assigned them a role as a buyer or a seller. Each market had 7 buyers and 7 sellers. These
roles were fixed for the duration of each session to enable subjects to gain experience with
a particular role. The subjects then repeatedly played a game with one another that approximated the model presented in the previous section.
Specifically, subjects participated in a total of 20 markets per session. Each market consisted of two stages. The first stage was a payment choice stage. In this first stage, each
buyer was endowed with 7 experimental money (EM) units and had to decide how to allocate his/her 7 EM between the two payment methods. To avoid any biases due to framing
effects, we used neutral language throughout, referring to cash and e-money as “payment 1”
and “payment 2,” respectively. Thus, in the first stage, buyers allocated their 7 EM between
payment 1 and payment 2, with only integer allocation amounts allowed, e.g., 3 EM in the
form of payment 1 and 4 EM in the form of payment 2. Each seller was endowed with 7 units
of goods. Sellers were required to accept payment 1 (cash) but had to decide in the first stage
whether or not to accept payment 2 (e-money) for that market. Sellers who decided to accept
payment 2 had to pay a one-time fee of T EM that enabled them to accept payment 2 in all
trading rounds of that market. As explained below, T is related to the costs described in the
model and is our main treatment variable.
During the first stage, subjects were also asked to forecast other participants’ payment
choices for that market. We elicited these forecasts because we wanted to better understand
subjects’ decision-making process. Specifically, buyers were asked to forecast how many
of the seven sellers would choose to accept payment 2 in that market. Sellers were asked
to supply two forecasts: (1) the average amount of EM that all seven buyers would allocate
to payment 2, and (2) how many of the other six sellers would choose to accept payment
2 in that market. Forecasts were incentivized; subjects earned 0.5 EM per correct forecast
in addition to their earnings from buying/selling goods (also in EM). The seller’s forecast
of the average of all buyers’ payment 2 allocations is counted as correct if it lies within
1 of the realized value. The other two forecasts were counted as correct only if they
precisely equalled the realized value. Note that no participant observed any other seller or
buyer’s choices or forecasts in this first stage, that is, all choices and forecasts were private
information and were made simultaneously.
Following completion of the first stage of each market play immediately proceeded to the
second trading stage of the market which consisted of a sequence of seven trading rounds. In
these seven rounds, each subject anonymously met with each of the seven subjects who were
11
in the opposite role to themselves, sequentially and in a random order. In each meeting the
buyer and seller tried to trade one unit of good for one unit of payment (recall that the terms
of trade in our model are fixed). Specifically, when each buyer meets each seller (and not
earlier), the buyer learned whether the seller accepted payment 2 or not, and then the buyer
alone decided which payment method to use, conditional on the buyer’s remaining balance for
that market of either payment 1 or payment 2; sellers were passive in these trading rounds,
simply accepting payment 1 from the buyer or payment 2 if the seller had paid the onetime fee to accept payment 2 in that market. Thus, provided that a buyer had some amount
of a payment type that the seller accepted, trade would be successful. For each successful
transaction, both parties to the trade earned 1 EM less some transaction costs for the trading
round, where the transaction costs depended on whether payment 1 or 2 was used as detailed
below.8 Notice that the only instances in which a transaction could not take place (was never
successful) were those in which the buyer had only payment 2 and the seller did not accept
payment 2. In those cases no trade could take place and both parties earned 0 EM for the
trading round.
Following completion of the seven trading rounds of the second stage of a market, that
market was over. Provided that the 20th market had not yet been completed, play then proceeded to a new two-stage market wherein buyers and sellers had to once again make payment
choices and forecasts in the first stage and then engage in 7 rounds of trading behavior in the
second stage. Buyers were free to change their payment allocations and sellers were free
to change their payment 2 acceptance decisions from market to market but only in the first
stage of the market; the choices made in this first stage were then in effect for all 7 rounds
of the second trading stage of the market that followed. Thus in total there were 20 markets
involving 7 trading rounds each or 140 trading rounds in total per session. In each trading round subjects could as much as 1 EM less transaction costs and in the first stage, they
could earn 0.5 EM per correct forecast. Following completion of the 20th market subjects
were paid their earnings in EM at the rate of 1 EM= 0.15 cents and in addition were paid a
$7 show-up payment. Each session lasted for about two hours. The average earnings were
between $15 25.
To facilitate decision making, we provided subjects with two types of information in stage
8
Note that while buyers were endowed with 7 EM at the start of each market, this endowment had to be
allocated between payment 1 and payment 2 for the buyers to actually earn EM in each of the 7 trading rounds
of the second stage of the market. Buyers could not choose to refuse to engage in trade and redeem their
endowment of EM. Unused payment allocations had no redemption value. Further, EM earnings from each
market were not transferrable to subsequent markets; instead these earnings were recorded and paid out only at
the end of the experiment following the completion of the 20th market. Thus, buyers started each new market
with exactly 7 EM and had to make payment allocations and payment choices anew in each market in order to
earn EM in that market.
12
1, at the same time that buyers were asked to make their payment allocation choices and
sellers were asked to make their payment 2 acceptance decisions and both types had to form
forecasts as described above. The first piece of information provided to subjects consisted
of payoff tables. The buyer’s payoff table reported the buyer’s market earnings if the buyer
allocated between 0~7 EM to payment 2 (and his/her remaining EM to payment 1) and if 0~7
sellers accepted payment 2. The seller’s payoff table reported the expected market earnings
the seller could get from the two options (accept/reject payment 2) in cases where all buyers
choose to allocate between 0~7 EM to payment 2, and where 0~6 of the other six sellers
choose to accept payment 2.9 In addition to these payoff tables, sellers also had access to a
“what if” calculator that computed their expected earnings in asymmetric cases where the 7
buyers made different payment allocation choices. The second piece of information that we
provided subjects in stage 1 (following the completion of the first market and every market
thereafter) was a history of outcomes in all past markets, including the subject’s payment
choices, the number of transactions using each of the two payment methods, the number
of no-trade meetings, market earnings from trading, and the number of correct forecasts.
Beginning with market 2, we also reported an aggregate market-level statistic, the number of
sellers who had chosen to accept payment 2 in prior market. We provided this information
so that sellers could learn about other sellers’ payment 2 acceptance decisions in the just
completed market; since all buyers visit all sellers and learn whether each seller accepts
payment 2 or not, buyers already had this piece of information by the end of each market.
By providing this information also to sellers, we made sure that both sides of the market had
symmetric information about seller choices.
We set the per transaction cost to be the same for all buyers and sellers, i.e., b = s =
= 0:5 for payment 1, and eb = es = e = 0:1 for payment 2. Thus, it was always the case
e
that
= 0:4 in all treatments of our experiment. Our only treatment variable was the
once per market fixed cost, T , that sellers had to pay to accept payment 2, which corresponds
to the parameter F in the model with a continuum of agents via the transformation T = 7F .
We chose three different values for this main treatment variable: T = 1:6, 2:8, and 3:5,
respectively.
Note that, given the lower transaction cost from using e-money, buyers are always better
off in the all e-money (all payment 2) equilibrium relative to the all cash (all payment 1).
e
However, seller’s relative payoffs depend on the fixed cost, T . If T < 7(
), as in our
T = 1:6 treatment (and represented graphically in the left panel of Figure 1) then the seller’s
payoff (like the buyer’s payoff) is higher in the all e-money equilibrium than in the all cash
9
See the experimental instructions in the appendix which include the payoff tables for both the buyers and
the sellers.
13
e
equilibrium. If T > 7(
), as in our T = 3:5 treatment (and represented graphically in
the right panel of Figure 1), then the seller’s payoff is lower in the all e-money equilibrium
e
than in the all cash equilibrium. Finally, if T = 7(
), as in our T = 2:8 treatment, then
the seller’s payoffs are the same in both the cash and e-money equilibria.
In terms of theoretical predictions, under our three different treatment conditions, there
always exist two symmetric pure-strategy Nash equilibria, one in which no seller accepts
payment 2 and all buyers allocate all of their endowment to payment 1 (the all cash equilibrium) and another equilibrium in which all sellers accept payment 2 and all buyers allocate
all of their endowment to payment 2 (the all e-money or card equilibrium). Given our parameterization, the symmetric payment 2 equilibrium is always the one the maximizes social
welfare.
While our focus is on symmetric equilibria, we note that there may also exist asymmetric
equilibria where some fraction of sellers accept payment 2 while the remaining fraction does
not, and buyers adjust their portfolios of payment 2 and 1 so as to perfectly match this distribution of seller choices. These asymmetric equilibria are always present in the T = 2:8;
e
treatment where T = 7(
). In particular, any outcome where ms 2 f1; 2; :::; 6g sellers
accept payment 2, while 7 ms do not, and all buyers allocate the same ms units of their
endowment to payment 2 and 7 ms to payment 1 comprises an asymmetric equilibrium
for this treatment. Thus, these asymmetric equilibrium characterize dual payment outcomes,
where both cash and e-money are used to make transactions. Due to our use of a finite
population size of 14 subjects there are also some asymmetric pure strategy equilibria of this
same variety in the T = 1:6 and T = 3:5 treatments, but these asymmetric equilibria are
fewer in number than in the T = 2:8 treatment and they would disappear completely as the
population size got larger and we approached the continuum of the theory, whereas the set
of asymmetric equilibria in the T = 2:8 case would continue to grow and would eventually
reach a continuum as in the theoretical model.
This multiplicity of equilibrium possibilities motivates our experimental study; equilibrium selection is clearly an empirical question that our experiment can help to address. We
hypothesize that, as the transaction cost to sellers of accepting payment 2 increases from
T = 1:6 to T = 2:8 and on up to T = 3:5, coordination on the e-money equilibrium will
become less likely and coordination on the cash equilibrium will become more likely, however, this remains an empirical question. In addition, we are interested in understanding the
dynamic process of equilibrium selection in terms of the evolution of subjects’ beliefs and
choices.
The experiment was computerized and programed using the z-Tree software (Fischbacher,
14
2007). At the beginning each session, each subject is assigned a computer terminal, and written instructions were handed out explaining the payoffs and objectives for both buyers and
sellers. See the appendix for example instructions used in the experiment. The instructor read
these instructions aloud in an effort to make the rules of the game public knowledge. Subjects could ask questions in private and were required to successfully complete a quiz to check
their comprehension of the instructions prior to the start of the first market. Communication
among subjects was prohibited during the experiment.
We have four sessions for each of our three treatment conditions, T = 1:6, T = 2:8 and
T = 3:5. As each session involved 14 subjects with no prior experience participating in our
study, we have data from 4 3 14 = 168 subjects. The experiment was conducted at Simon
Fraser University (SFU), Burnaby, Canada and at the University of California, Irvine (UCI),
USA using undergraduate student subjects. Specifically, two sessions of each of our three
treatments (one-half of all sessions) were run at SFU and UCI, respectively. Despite our use
of two different subject pools, we did not find significant differences in either buyer or seller
behavior across these two locations as we show later in the paper.
4
Aggregate Experimental Results
In this section, we present and discuss our experimental results at the aggregate level. The
next section will address individual behavior.
Figures 2 to 3 show the evolution of behavior over time in each of the four sessions of
our three treatments. In all of these figures, the horizontal axis indicates the number of the
market, running from 1 to 20. Figures 2abc, display time series on payment choices and
transaction methods in two separate panels for each session. In the first panel the series labeled "BuyerPay2," shows the percentage of the endowment that all buyers allocated to payment 2 averaged across the seven buyers of each session. In this same panel, the percentage
of sellers accepting payment 2 is indicated by the series labeled "SellerAccept." The second
panel of Figures 2abc show three time series: 1) the frequency of meetings that resulted in
transactions using payment 1 labeled as "Pay1," 2) the frequency of meetings that resulted
in transactions using payment 2 labeled as "Pay2," and 3) circles indicating the frequency of
no-trade meetings labeled as "NoTrade." Figures 3abc also show two panels for each session
of the three treatments. The first panel illustrates the average buyer, seller and both buyer
and seller (all) payoffs as a percentage of the payoffs that could be earned in the symmetric
pure strategy equilibrium where all buyers use payment 2 and all sellers accept payment 2
– the all-payment-2 equilibrium. The second panel is similar but shows the average buyer,
15
seller and combined buyer and seller (all) payoffs as a percentage of the payoffs that would
be earned in the symmetric pure strategy equilibrium where all buyers use payment 1 and all
sellers refuse to accept payment 2, – the all-payment-1 equilibrium.
Tables 1 and 2 display various statistics for each of the 12 sessions of the experiment
using the same time series data that is depicted in Figures 2-3. For each statistic, we show
the treatment-level average in bold face. Table 1 provides five statistics on payment choice
and usage and transactions. The first part (rows 1 to 10) of Table 1 reports the percentage of
endowment allocated to payment 2 averaged across the seven buyers and the percentage of the
seven sellers accepting payment 2. In particular, we report the session mean, minimum and
maximum, and the mean in the first and the last markets for these two statistics. The second
part of Table 1 (rows 16 to 25) shows the percentage of meetings that resulted in trading with
payment 1, trading with payment 2 and no trading in a market. Again, we provide the session
mean, minimum and maximum, and the mean value in the first and last markets for these
statistics. Table 2 provides the same set of statistics on payoff efficiency for buyers, sellers
or both ("all") relative to the all-payment-2 or all-payment-1 equilibrium benchmarks. We
use the data reported in Tables 1-2 (and depicted over time in Figures 2-3) to characterize the
following aggregate findings.
Finding 1 Across the three treatments, as T is increased from 1.6 to 2.8 to 3.5, there are
significant decreases in the buyer’s choice of payment 2, the seller’s acceptance of payment
2, and successful transactions involving payment 2.
Support for Finding 1 comes from Table 3 which reports results from a Wilcoxon ranksum test of treatment differences using session level averages (four per treatment). The test
results indicate that the buyer’s choice of payment 2 (BuyerChoice) the seller’s acceptance
of payment 2 (SellerAccept) and successful payment 2 transactions (Pay2Meetings) are significantly higher in the T = 1:6 treatment as compared with either the T = 2:8 or T = 3:5
treatments (p < :05). Further, these same means are higher in the T = 2:8 treatment as
compared with the T = 3:5 treatment. Importantly, Table 3 also indicates that initially, using
means from just the first market of each session (“First market”) there are no differences in
these three mean statistics across our three treatments, indicating that all session started out
with roughly similar initial conditions and the choices of buyers and sellers then evolved over
time to yield the differences summarized in Finding 1. The next three findings summarize
aggregate behavior in each of the three experimental treatments.
Finding 2 When T = 1:6, the experimental economies converge to (or nearly converge to)
the all-payment-2 equilibrium.
16
Support for Finding 2 comes from Table 1 and Figure 2a which report on the evolution
of behavior in the four sessions of the T = 1:6 treatment. In all four sessions, subjects
move over time in the direction of the payment 2 equilibrium, which in this case represents a
strict Pareto improvement for both sides. Furthermore, sellers do not suffer much loss from
investing on the fixed cost: they can recover the fee if each buyer allocates just 4 EM (57%) or
more to payment 2. As a result, sellers maintain high levels of acceptance, and this steadily
high acceptance rate encourages buyers to quickly catch up, which, in turn, reinforces the
incentives for acceptance of payment 2. As a result, toward the end of all four T = 1:6
treatment sessions, subjects have achieved convergence or near-convergence to the payment
2 equilibrium.
Finding 3 When T = 2:8, there is a mixture of outcomes consistent with the greater multiplicity of equilibria in this treatment.
Support for Finding 3 comes from Table 1 and Figure 2b which report on the evolution
of behavior in the four sessions of the T = 2:8 treatment. By contrast with the T = 1:6
treatment, the data reported in Table 1 and Figure 2b reveal a mixture of outcomes across
the four sessions of the T = 2:8 treatment. In particular, we observe that the experimental
economy either lingers in the middle ground between the all payment 1 and all payment 2
equilibria (sessions 1, 3 and 4) or appears to be very slowly converging toward the all payment 2 equilibrium (session 2). Recall that when T = 2:8; sellers are indifferent between the
two equilibria as their payoffs are the same in either equilibrium. Further any outcome where
all buyers allocate the same proportion of their endowment to payment 2 as the fraction of
sellers accepting payment 2 is always an asymmetric equilibrium in this setting. Generally we
observe that the fraction of sellers accepting payment 2 hovers above (with some volatility)
the fraction of endowment that buyers allocate to payment 2. If, over time, buyers increasingly insist on the new payment method 2, then sellers are likely to accommodate the buyers’
choices by accepting it; this seems to be the case in session 2. In the final market of session 2,
buyer’s payment 2 allocation averages 96%, and six out of seven sellers (86%) are accepting
payment 2. The number of payment 2 transactions increases from 55% in the first market to
84% in the last market. Note that compared with the T = 1:6 treatment sessions, the process
of convergence to the all payment 2 equilibrium is considerably slower, more erratic and incomplete. By contrast, in the other three sessions there is little to no evidence of convergence
toward either the all cash or all e-money equilibrium. In session 3 there is a small increase in
the number of transactions using payment 2 from 61% in the first market to 76% in the final
market, but the economy remains far away from the all payment 2 equilibrium, or any other
symmetric equilibrium. In session 1, the number of sellers accepting payment 2 fluctuates
17
between 2/7 (28.5%) and 5/7 (71.4%). Consequently, buyers are not willing to take the lead
by acquiring high payment 2 balances, fearing that they will not be able to trade in case some
sellers reject it. As a result, the buyer’s average payment 2 allocation and the number of
sellers accepting payment 2 average between 40-50% throughout the entire session, but there
is never coordination on the same rate. Finally, session 4 shows an upward trend in payment
2 usage in the first seven markets, but this trend abruptly halts thereafter, with the average
number of payment 2 transactions barely changing from market eight onward (the value fluctuates between 67% and 71%). This outcome represent near (but imperfect) convergence
to an interior equilibrium where approximately 5 out 7 sellers are accepting payment 2 and
buyers are allocating approximately 5 units of their 7 EM endowment to payment 2. This is
the closest instance we have to a dual payments equilibrium outcome.
Finding 4 When T = 3:5, the experimental economy slowly converges to the payment 1
equilibrium, or lingers in the middle ground between the two pure-strategy equilibria.
Support for Finding 4 comes from Table 1 and Figure 2c which report on the evolution of
behavior in the four sessions of the T = 3:5 treatment. When T = 3:5, sellers do better in
the payment 1 equilibrium as compared with the payment 2 equilibrium; by contrast buyers
always prefer the payment 2 equilibrium. Nevertheless, in each of the four sessions, more
than 50% of sellers start out in the first market accepting payment 2 perhaps fearing that they
will lose business in case where some buyers show up with only payment 2 remaining. With
experience sellers learn to resist accepting payment 2 and engage in a tug of war with buyers;
the average acceptance rate over all four sessions declines from 75% in the first market to
just 21% in the last market. Sellers appear to be winning this contest in sessions 1, 2 and 4,
pulling the economy back in the direction of the status quo, payment-1-only equilibrium. For
example, in session 1, the buyer’s payment 2 allocation falls from 65% in the first market to
an average of just 6% in the last market. On the seller’s side, 6 of 7 (86%) of sellers accept
payment 2 in market 1; by the end of the session, no seller is accepting payment 2. The
number of payment 1 transactions increases from 35% in the first market to 94% by the final,
20th market; over the same interval, payment 2 transactions fall from 63% to 0%. In session
4, the trend works in favor of the seller, but the speed of convergence is very slow; at the end
of the session, two sellers (29%) continue to accept payment 2 and 22% of transactions are
conducted in payment 2 (however it seems reasonable to conjecture that the economy would
get even closer to the payment 1 equilibrium if the session lasted more than 20 markets). In
session 3, the tug of war continues throughout the session and neither side is able to gain
the upper hand; in that session, the average buyer’s payment 2 allocation and the number of
sellers accepting payment 2 consistently fluctuates around 50%, but there is no coordination
18
on any asymmetric equilibrium in this setting.
An immediate implication of Findings 1-4 is the following:
Finding 5 Efficiency losses increase with increases in T .
Support for Finding 5 comes from Table 2 and Figures 3abc.
When T = 1:6, the economy quickly converges to the socially efficient payment 2 equilibrium. All four sessions achieve between 92 and 96% of the socially optimal (payment-2only) equilibrium payoffs, with the overall average being 94%. The efficiency measure falls
as T increases to 2:8, ranging from 75% to 85%, with a treatment average of 81%. Treatment T = 3:5 induces a further decrease in the efficiency measure, with efficiency measures
ranging from 72% to 75% percent and and a treatment average of 75%. As the fixed cost
increases, the economy moves further away from the efficient equilibrium. At the same time,
mis-coordination in payment choices becomes more severe, as manifested in the increasing
frequency of no-trade meetings, which averaged 0:8% for T = 1:6, 5:6% for T = 2:8 and
8:6% for T = 3:5.
The picture is unchanged if we consider earnings relative to the status quo where only
payment 1 is used. The percentage of subjects’ earnings relative to the payment 1 equilibrium
level serves as a measure of the benefit of introducing the new payment method, payment 2.10
As revealed in the bottom half of Table 2, there are significant positive welfare benefits to the
introduction of payment 2 when T = 1:6, moderate benefits when T = 2:8, but almost
no benefit when T = 3:5. Disaggregating by role, Table 2 further reveals that in all three
treatments buyers benefit relative to the payment 1 equilibrium, while sellers only benefit in
T = 1:6 treatment and sellers suffer in the other two treatments relative to the payment 1
equilibrium benchmark. The latter finding is summarized as follows.
Finding 6 Sellers may choose to accept the new payment method (payment 2) even if doing
so reduces their payoffs relative to the status quo where only payment 1 is used.
Merchants often complain about high costs associated with accepting electronic payments
but feel obliged to accept those costly payments for fear of upsetting or losing their customers.11 We observe a similar pattern in our experiment. For instance, when T = 3:5,
10
Since payment 1, representing cash, has to be accepted by legal restriction, it is natural to think of payment
2 as the new and competing payment method.
11
Some evidence in support of this claim comes form Evans (2011, p. vi) who observes that “...the US
Congress passed legislation in 2010 that required the Federal Reserve Board to regulate debit card interchange
fees; the Reserve Bank of Australia decided to regulate credit card interchange fees in 2002 after concluding that
a market failure had resulted in merchants paying fees that were too high; and in 2007 the European Commission
ruled that MasterCard’s interchange fees violated the EU’s antitrust laws.”
19
although sellers understand that they will lose relative to the status quo if the economy moves
to the payment 2 equilibrium most sellers still begin the session accepting the new payment
method (the average frequency of sellers accepting payment 2 in the first market is 70% in
treatment T = 2:8 and 76% treatment T = 3:5). Figures 3b and 3c reveal that in all sessions
with T = 2:8 and 3:5, the seller’s average payoff is always below 3:5, the payoff that they
would earn were the economy staying in the payment 1 equilibrium.
5
Individual Experimental Results
In this section we explore in further detail individual decision-making in our experiment.
In particular, we first examine how the portfolio decisions of buyers depend on outcomes
in past markets and their beliefs for the current market. We then do the same for seller’s
decisions to accept or not accept payment 2.
Table 4 reports regression estimates from a linear, random-effects model of the buyer’s
payment 2 allocation choice (card choice) where the random effect is at the individual buyer
level. Specifically, the dependent variable is the percentage of the buyer’s endowment that
s/he allocated to payment 2 (card choice %). The two main explanatory variables are: mktAcceptL(%), representing the percentage of sellers accepting payment 2 in the last market,
and bBelief%, the buyer’s own incentivized belief as to the number of sellers who would be
accepting payment 2 in the current market.12 Additional explanatory variables are the market
number, 1,2...20 (“market”) to capture learning effects, a location dummy (“location") equal
to 1 if the data were collected at SFU and two further treatment dummies, T 16 and T 35; to
pick up treatment level effects from the T = 1:6 and T = 3:5 treatments respectively (the
baseline treatment is thus T = 2:8).
The first column of Table 4 reports results using the pooled data from the entire experiment using all six of these explanatory variables. The results indicate that all variables except
the location dummy are statistically significant; the latter finding tells us that on the buyer
side there was no significant difference in buyer behavior between SFU and UCI and thus
rationalizes our pooling of the data from these two populations. We note that among the statistically significant explanatory variables, all but the coefficient on the T35 dummy variable
have positive coefficients.
The interpretation of these results is straightforward; buyers increase their allocation to
12
Buyers learned the percentage of sellers accepting payment 2 in the prior market because they visited all
7 sellers and were informed in every instance whether or not the seller accepted payment 2. In addition, we
reported information on the percentage of sellers who accepted payment 2 in all prior markets on buyers’ stage
1 portfolio allocation screen. Thus buyers had ready access to the statisitc mktAcceptL.
20
payment 2 the greater is: past market seller acceptance of payment 2; buyers’ beliefs about
seller’s acceptance of payment 2 in the current market; the market number and if T = 1:6. If
T = 3:5, there is a significant drop in buyers’ allocations to payment 2, as buyers anticipate
the consequences of higher seller fixed costs for seller acceptance of payment 2. The last
three columns of Table 4 report on the same regression model specifications but using the
data separately from each of the three treatments (thus omitting the T16 and T35 dummies).
As these last three columns reveal, buyers’ allocation of their endowment to payment 2 is
again increasing in both the past market percentage of sellers accepting payment 2 and in
buyers’ beliefs about the percentage of sellers who will accept of payment 2 in the current
market. However, the coefficient on the market variable ranges from a positive and significant
0:699 when T = 1:6 to 0:197 when T = 2:8 to a negative and significant 0:404 when
T = 3:5, which indicates that buyer behavior is consistent with Finding 1. Notice further that
the location dummy variable is significantly positive for T = 1:6 and T = 2:8 treatments,
but negative in the T = 3:5 treatment, which indicates that while there were differences in
individual buyer behavior across the two locations at the treatment level, overall, across all
three treatments there is no systematic bias (all positive or all negative) in buyers’ allocations
of endowment to payment 2 (as confirmed again by the insignificance of the location dummy
using the pooled data).
As each seller’s decision to accept payment 2 or not is a binary choice, Table 5 reports on a
random effects probit regression analysis of the factors affecting individual sellers’ payment 2
acceptance decisions where the random effect is at the individual seller level. For the pooled
data analysis from all three treatments (first column of Table 5) we consider nine explanatory
variables, the last four of which are the same ones that were used in the regression analysis
reported in Table 4. The five other explanatory variables are: sOtherAcceptL(%), which is
the percentage of sellers who accepted payment 2 in the previous market (this information
was only revealed to sellers at the start of stage 1 of each new market when they had to
make a payment 2 acceptance choice and not earlier); sAcceptL*sCardDealL(%), which is
the percentage of transactions the seller succeeded in conducting using payment 2 in the
previous market conditional on his having accepted payment 2 in that previous market, i.e. if
sAcceptL=1; (1-sAcceptL)*sNoDealL(%) is the percentage of no trade outcomes the seller
encountered in the previous market conditional on his having refused to accept payment
2 in that previous market, i.e. if sAcceptL=0; and sBeliefB and sBeliefS are the seller’s
incentivized beliefs about the buyers’ average allocation of endowment to payment 2 and of
the percentage of the other 6 sellers (excluding themselves) who would accept payment 2,
respectively, in the current market.
Table 5 reports marginal effects of these explanatory variables on the seller’s probability
21
of accepting payment 2. For the pooled data estimates (first column of Table 5), we observe
that the percentage of other sellers accepting payment 2 in the prior market has no statistically
significant effect on a seller’s current market decision to accept payment 2. However, sellers
do respond to their own prior market experience from accepting or not accepting payment 2.
In particular, sellers who accepted payment 2 in the prior market are more likely to accept
it again, the higher were the percentage of transactions they completed using payment 2 in
that prior market. Sellers who did not accept payment 2 in the prior market are more likely to
accept it in the current market, the larger the number of no trade transactions they experienced
in that prior market.13 The disaggregated treatment-level regression estimates reported on in
the last three columns of Table 5 reveal that these latter two effects are coming mainly from
the T = 3:5 treatment; in that treatment, sellers face the highest fixed cost for adopting
payment 2 so they more carefully pay attention to the payoff consequences of prior payment
2 acceptance or non-acceptance decisions in this treatment relative to the other two where the
fixed costs of accepting payment 2 were lower.
We further observe that sellers’ willingness to accept payment 2 is positively affected by
their beliefs about the percentage of buyers’ endowment that would be allocated to payment
2 in the current market and by their beliefs about the percentage of other sellers’ who would
accept payment 2 in the current market. The latter finding holds both for the pooled data and
for the individual treatment specifications. The market variable is not statistically significant
in the pooled seller regression, though it is negative and significant for the T = 2:8 treatment
alone; sellers in that treatment were more likely to accept payment 2 with experience. There
is again an absence of any location effect on the seller side as evidenced by the insignificant estimate on the location dummy variable both in the pooled data and in the individual
treatment specifications. However, there is again a strong treatment effect as indicated by the
positive and negative coefficients on the T16 and T35 dummies, respectively; sellers in the
T = 1:6 treatment were more likely to accept payment 2 while those in the T = 3:5 treatment
were more likely to reject payment 2, all relative to the T = 2:8 treatment baseline.
Summarizing the results of this section, we have found that past market outcomes and
beliefs matter for both buyer and seller behavior in a manner that is consistent with our
theory. Buyers pay attention to the prior distribution of sellers accepting payment 2 when
making portfolio allocations and sellers condition their payment 2 acceptance decisions on
their own prior experiences with accepting or not accepting payment 2. Buyers act on their
beliefs about the current distribution of sellers accepting payment 2 and sellers act on their
beliefs about the current percentage of payment 2 held by buyers and on the current decisions
13
A no-trade outcome can only occur in the case where the seller does not accept payment 2 and the buyer
has only payment 2 in his/her portfolio. Sellers must always accept payment 1 (cash).
22
of other sellers to accept payment 2. We have also confirmed the strong treatment effects we
report on earlier in 1. Finally, we have not found evidence for any strong location effects in
the decisions made by buyers or sellers, despite the fact that we conducted our experiment
using student subjects at two different universities, SFU and UCI.
6
Conclusion and Directions for Future Work
We have developed a simple model to understand factors that may contribute to the adoption of a new payment method when there already exists a payment method that all sellers
accept. Specifically, we have isolated two factors: network adoption effects and seller transaction costs as determinants of whether or not the new payment method can take the place
of the existing payment system. As our model admits multiple equilibria for a wide set of
parameter values, we have chosen to study the behavior of human subjects placed in a simple
version of that model, incentivize them to make decisions in accordance with the theory and
give them ample opportunities to learn how to make choices in that environment.
Our model/experiment involves a two stage game. In the first stage, sellers decide whether
or not to accept the new payment method; they must always continue to accept the old payment method. Simultaneously in this first stage, buyers decide how to allocate their endowment between the two payment methods without knowing of seller acceptance decisions. To
use the new payment method, the seller has to pay a fixed cost to accept it, but the new payment method saves on per transaction costs for both buyers and sellers. Due to the network
effect, the status quo where the existing payment is used and complete adoption of the new
payment method are both equilibria.
Our main treatment variable is the fixed cost to accept the new payment method. We find
that the new payment method will take off if the fixed cost is low so that both sides benefit
by switching to the new payment method. If the fixed cost is high such that the seller endures
a loss in the equilibrium where the new payment method is used relative to the equilibrium
where it is not accepted, some sellers nevertheless respond by accepting the new payment
method initially, fearing to lose business, but they mostly eventually learn over time to resist
the new payment method and pull the economy back to the old payment method. If neither
side displays much willpower to move behavior toward one equilibrium or the other, then the
economy may linger in the middle ground between the two equilibria.
The framework we have developed in this paper can be used to analyze a series of new
questions. (1) What are the effects of subsidies and taxes on the adoption of a new payment
method? Consider schemes with balanced budgets where subsidies are financed by taxes.
23
Which schemes are more effective in promoting the new payment method? Charging sellers
to subsidize buyers or the other way around? (2) In the baseline model, we fix the terms of
trade. What will happen if we allow sellers to offer discounts or impose surcharges (i.e., passthroughs). Will sellers use discounts or surcharges? How will discounts or surcharges affect
the diffusion of the new payment method? (3) The baseline model assumes a single new payment method. It is of interest to further explore how the economy evolves if there is more than
one new payment method. Imagine the simplest case featuring two new payment methods
with the same setup cost. Due to the positive setup cost associated with each new payment
method, the most efficient allocation is to use only one new method. Competition between
the two new payment methods may make it difficult for agents to coordinate on a single new
payment method. As a result, it is possible that neither payment method will take off, or it
may take a longer time for the economy to converge to the efficient equilibria. (4) What are
the desirable features of the existing payment method, i.e., cash that the new payment method
needs to retain in order to make it more competitive? Anonymity? Robustness to network
breakdowns? What is the effect of an incidence of identity theft? Suppose the economy has
arrived at the equilibrium with the new payment method. Can the economy revert to the cash
only equilibrium and if so, under what circumstances? (5) Suppose consumers receive their
income in the form of either the old or the new payment method, and assume there is a small
cost of portfolio adjustment at the beginning of each trading period. Do these changes matter for the equilibrium that is selected? (6) The baseline model assumes that sellers always
accept the old payment method. What would happen if we relax this assumption and allow
sellers to decide whether to accept the old payment method? (7) The baseline model assumes
that consumers cannot convert between the old and the new payment methods after the initial
portfolio choice. As a result, if the consumer has only the new payment method but the seller
does not accept it, there will be no trade. In such instances, we could modify the model to
allow for conversion opportunities subject to a cost. Note that this new specification does not
change the equilibrium outcomes. In equilibrium, consumers’ portfolio choice will be fully
consistent with the seller’s acceptance pattern, and there is no need to convert. However, the
conversion cost(s) may affect the speed of convergence to equilibria. We believe that all of
these extensions of our model are worth examining theoretically and/or experimentally but
we must leave such an analysis to future research.
24
References
Arango, Carlos, Kim P. Huynh, and Leonard Sabetti (2011), “How do you pay? The role of
incentives at the point-of-sale.” Bank of Canada working paper 2011-23.
Arango, Carlos, and Varya Taylor (2008), “Merchants’ Costs of Accepting Means of Payment: Is Cash the Least Costly?” Bank of Canada Review, Winter 2008-2009, 17–25.
Bagnall, John, David Bounie, Kim P. Huynh, Anneke Kosse, Tobias Schmidt, Scott Schuh,
and Helmut Stix (2015), "Consumer cash usage: a cross-country comparison with payment diary survey data," Forthcoming in International Journal of Central Banking.
Camera, Gabriele, Marco Casari and Stefania Bortolotti (2015), “An experiment on retail
payment systems.” Forthcoming in Journal of Money, Credit and Banking.
Chiu, Jonathan, and Tsz-Nga Wong (2014), “E-money: efficiency, stability and optimal
policy.” Bank of Canada working paper 2014-16.
Cohen, Michael, and Marc Rysman (2013), “Payment choice with consumer panel data,”
Working Paper 13-6, Federal Reserve Bank of Boston.
Evans, David S. (2011), "Interchange fees: the economics and regulation of what merchants
pay for cards," Boston: Competition Policy International.
Fischbacher, Urs (2007), “z-Tree: Zurich toolbox for ready-made economic experiments.”
Experimental Economics 10(2), 171-178.
Klee, Elizabeth (2008), “How people pay: Evidence from grocery store data,” Journal of
Monetary Economics 55, 526-541.
Koulayev, Sergei, Marc Rysman, Scott Schuh, and Joanna Stavins (2012), “Explaining
adoption and use of payment instruments by U.S. consumers.” Working Paper 12-14,
Federal Reserve Bank of Boston.
McAndrews, James, and Zhu Wang (2012), “The economics of two-sided payment card
markets: pricing, adoption and usage.” Working Paper 12-06, Federal Reserve Bank of
Richmond.
Rochet, Jean-Charles, and Jean Tirole (2002), “Cooperation among competitors: some economics of payment card associations.” RAND Journal of Economics 33(4), 549-570.
Schuh, Scott, and Joanna Stavins (2013), “How consumers pay: adoption and use of payments.” Accounting and Finance Research 2(2), 1-21.
25
Von Kalckreuth, Ulf, Tobias Schmidt and Helmut Stix (2014), “Choosing and using payment
instruments: evidence from German microdata.” Empirical Economics 46(3), 10191055.
Wright, Julian (2003), “Optimal card payment systems.” European Economic Review 47,
587-612.
26
Figure 2a: Payment Choice and Usage T=1.6
Session 1
Session 2
Session 3
Session 4
Notes. (1) Horizontal axis: market. (2) There are two figures for each session. The first shows payment choices: the red dashed line represents the percentage of
money allocated to payment 2, averaged across the seven buyers; the blue solid line represents the percentage of the seven sellers accepting payment 2. The second
figure describes payment usage. The solid blue line is the percentage of meetings using payment 2; the dashed blue line is the percentage of meetings using
payment 1; red circles are the percentage of meetings where no trade takes place.
Figure 2b: Payment Choice and Usage T=2.8
Session 1
Session 2
Session 3
Session 4
Notes. (1) Horizontal axis: market. (2) There are two figures for each session. The first shows payment choices: the red dashed line represents the percentage of
money allocated to payment 2, averaged across the seven buyers; the blue solid line represents the percentage of the seven sellers accepting payment 2. The second
figure describes payment usage. The solid blue line is the percentage of meetings using payment 2; the dashed blue line is the percentage of meetings using
payment 1; red circles are the percentage of meetings where no trade takes place.
Figure 2c: Payment Choice and Usage T=3.5
Session 1
Session 2
Session 3
Session 4
Notes. (1) Horizontal axis: market. (2) There are two figures for each session. The first shows payment choices: the red dashed line represents the percentage of
money allocated to payment 2, averaged across the seven buyers; the blue solid line represents the percentage of the seven sellers accepting payment 2. The second
figure describes payment usage. The solid blue line is the percentage of meetings using payment 2; the dashed blue line is the percentage of meetings using
payment 1; red circles are the percentage of meetings where no trade takes place.
Figure 3a: Efficiency T=1.6
Session 1
Session 2
Session 3
Session 4
Notes. (1) Horizontal axis: market. (2) There are two figures for each session. The first figure measures efficiency using payoffs in the payment 2 equilibrium as
the benchmark (100%). The blue solid line represents sellers, the red dashed line represents buyers and the black line marked with circles represents buyers and
sellers together. The second figure measures efficiency using payoffs in the payment 1 equilibrium as the benchmark.
Figure 3b: Efficiency T=2.8
Session 1
Session 2
Session 3
Session 4
Notes. (1) Horizontal axis: market. (2) There are two figures for each session. The first figure measures efficiency using payoffs in the payment 2 equilibrium as
the benchmark (100%). The blue solid line represents sellers, the red dashed line represents buyers and the black line marked with circles represents buyers and
sellers together. The second figure measures efficiency using payoffs in the payment 1 equilibrium as the benchmark.
Figure 3c: Efficiency T=3.5
Session 1
Session 2
Session 3
Session 4
Notes. (1) Horizontal axis: market. (2) There are two figures for each session. The first figure measures efficiency using payoffs in the payment 2 equilibrium as
the benchmark (100%). The blue solid line represents sellers, the red dashed line represents buyers and the black line marked with circles represents buyers and
sellers together. The second figure measures efficiency using payoffs in the payment 1 equilibrium as the benchmark.
Table 1: Payment choice and usage
Treatment
Session
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
session mean
% of money allocated to session min
payment 2
session max
first market
last market
session mean
% of sellers
session min
accepting
session max
payment 2
first market
last market
session mean
% of meetings
session min
using payment 2
session max
first market
last market
session mean
% of meetings
session min
using payment 1
session max
first market
last market
session mean
% of meetings
session min
with no trade
session max
first market
last market
1
85
53
100
53
98
98
86
100
100
100
84
53
100
53
98
15
0
47
47
2
1
0
12
0
0
2
93
53
100
53
100
99
86
100
86
100
92
53
100
53
100
7
0
47
47
0
1
0
14
0
0
T=1.6
3
88
65
100
67
100
93
71
100
71
100
86
61
100
63
100
12
0
35
33
0
2
0
14
4
0
4
94
63
100
63
100
99
86
100
100
100
93
63
100
63
100
6
0
37
37
0
1
0
12
0
0
all
90
59
100
59
99
97
82
100
89
100
89
58
100
58
99
10
0
41
41
1
1
0
13
1
0
1
44
37
57
57
43
46
29
71
29
29
38
27
49
27
29
56
43
63
43
57
6
0
31
31
14
2
74
55
96
73
96
79
29
100
57
86
67
27
92
55
84
26
4
45
27
4
7
0
43
18
12
T=2.8
3
68
61
80
61
80
74
57
100
86
86
62
55
76
61
76
32
20
39
39
20
6
0
22
0
4
4
67
49
76
49
69
72
43
100
100
71
64
43
71
49
69
33
24
51
51
31
3
0
22
0
0
all
63
51
77
60
72
68
39
93
68
68
58
38
72
48
64
37
23
49
40
28
6
0
30
12
8
1
38
6
71
65
6
33
0
86
86
0
29
0
63
63
0
62
29
94
35
94
9
0
22
2
6
2
33
14
65
49
16
31
0
86
86
14
23
0
53
49
12
67
35
86
51
84
9
0
51
0
4
T=3.5
3
53
39
71
39
43
54
29
86
71
43
45
27
59
39
39
47
29
61
61
57
8
0
31
0
4
4
36
22
57
45
22
33
14
71
57
29
28
14
49
41
22
64
43
78
55
78
8
0
43
4
0
all
40
20
66
49
22
38
11
82
75
21
31
10
56
48
18
60
34
80
51
78
9
0
37
2
4
Table 2: Efficiency
Part 1: payment 2 equilibrium as benchmark
Treatment
Session
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
buyer
seller
all
session mean
session min
session max
first market
last market
session mean
session min
session max
first market
last market
session mean
session min
session max
first market
last market
1
93
79
100
79
99
91
72
100
72
99
92
76
100
76
99
2
96
79
100
79
100
95
73
100
77
100
96
77
100
78
100
T=1.6
3
93
75
100
81
100
92
76
100
85
100
92
75
100
83
100
4
96
84
100
84
100
95
78
100
78
100
96
81
100
81
100
167
142
180
142
178
123
102
134
102
133
144
120
157
120
156
173
142
180
142
180
129
104
134
105
134
151
121
157
123
157
167
135
180
147
180
124
101
134
111
134
145
118
157
130
157
173
151
180
151
180
129
109
134
109
134
151
128
157
128
157
all
94
79
100
81
100
93
75
100
78
100
94
77
100
80
100
1
69
50
77
50
60
87
68
95
68
86
75
57
82
57
69
2
82
44
96
70
86
84
56
93
80
86
82
48
95
73
86
124
91
139
91
109
87
68
95
68
86
105
79
114
79
97
147
78
173
126
155
84
56
93
80
86
115
67
133
103
120
T=2.8
3
4
80
82
68
62
89
87
83
77
87
86
84
91
69
59
94
100
80
59
88
98
81
85
72
68
87
92
82
71
87
91
all
78
56
88
70
80
86
63
96
72
89
81
61
89
71
83
1
64
43
83
83
52
102
79
121
79
117
75
60
87
81
72
2
61
27
77
77
59
99
61
114
67
114
72
38
81
74
76
T=3.5
3
4
71
64
51
38
82
76
73
71
71
66
91
102
63
68
105 118
74
89
105 112
77
75
60
47
86
84
73
77
81
80
all
65
40
80
76
62
98
68
115
77
112
75
51
84
76
77
141
101
158
126
144
86
63
96
72
89
114
85
125
99
117
114
78
149
149
94
82
63
97
63
94
98
78
113
106
94
110
49
139
139
106
79
49
91
53
91
94
49
105
96
99
127
92
147
131
127
73
50
84
60
84
100
78
112
95
106
116
72
143
137
111
79
54
92
62
90
98
66
110
99
100
Part 2: payment 1 equilibrium as benchmark
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
buyer
seller
all
session mean
session min
session max
first market
last market
session mean
session min
session max
first market
last market
session mean
session min
session max
first market
last market
170
143
180
146
180
126
104
134
107
134
148
122
157
125
157
144
123
160
149
156
84
69
94
80
88
114
100
122
115
122
148
112
157
139
156
91
59
100
59
98
120
95
129
99
127
115
69
137
129
118
82
54
94
71
89
98
61
109
100
104
Table 3: Rank-sum Test – Treatment Effect
Rank-sum
group 1
Rank-sum
group 2
z-value
p-value
Session average
BuyerChoice
SellerAccept
Pay2Meetings
26
26
26
10
10
10
2.309
2.323
2.309
0.021
0.020
0.021
First market
BuyerChoice
SellerAccept
Pay2Meetings
18
21.5
22
18
14.5
14
0
1.042
1.169
1
0.298
0.243
Session average
BuyerChoice
SellerAccept
Pay2Meetings
25
25
25
11
11
11
2.021
2.033
2.021
0.043
0.042
0.043
First market
BuyerChoice
SellerAccept
Pay2Meetings
22.5
18.5
18.5
13.5
17.5
17.5
1.307
0.149
0.145
0.191
0.882
0.885
Session average
BuyerChoice
SellerAccept
Pay2Meetings
26
26
26
10
10
10
2.309
2.337
2.309
0.021
0.019
0.021
First market
BuyerChoice
SellerAccept
Pay2Meetings
23
22.5
23
13
13.5
13
1.452
1.348
1.488
0.147
0.178
0.137
T=1.6 versus T=2.8
T=2.8 versus T=3.5
T=1.6 versus T=3.5
Note: combined sample size for each test is 8
Table 4 : Buyer payment 2 choice (%) with random effects
Independent variables
MktAccceptL(%)
bBelief(%)
market
location (SFU=1;UCI=0)
T16
T35
No. of obs.
Overall R2
statistics pooled
Coef.
0.218
Std.Err.
0.021
t
10.33
p
0.000
Coef.
0.593
Std.Err.
0.021
t
27.90
p
0.000
Coef.
0.129
Std.Err.
0.049
t
2.64
p
0.008
Coef.
1.787
Std.Err.
1.212
t
1.47
p
0.140
Coef.
5.174
Std.Err.
1.591
t
3.25
p
0.001
Coef.
-4.042
Std.Err.
1.578
t
-2.56
p
0.010
***
***
***
T=1.6
0.328
0.108
3.02
0.002
0.400
0.048
8.29
0
0.699
0.124
5.64
0.000
4.624
2.140
2.16
0.031
***
***
***
**
T=2.8
0.100
0.027
3.64
0.000
0.624
0.034
18.34
0.000
0.197
0.072
2.75
0.006
5.575
2.151
2.59
0.01
***
***
***
***
T=3.5
0.174
0.025
6.82
0.000
0.567
0.029
19.68
0.000
-0.404
0.086
-4.71
0.000
-3.731
1.823
-2.05
0.041
***
***
1596
0.825
Note
(1) *p-value<=0.1; **p-value<=0.05; *** p-value<=0.01
532
0.355
532
0.687
532
0.764
***
***
***
**
Table 5: Seller acceptance (1=accept, 0=reject), probit with random effects
Independent variables
statistics
pooled
T=1.6
T=2.8
T=3.5
sOtherAcceptL(%)
dy/dx
Std.Err.
t
p
dy/dx
Std.Err.
t
p
dy/dx
Std.Err.
t
p
dy/dx
Std.Err.
t
p
dy/dx
Std.Err.
t
p
dy/dx
Std.Err.
t
p
dy/dx
Std.Err.
t
p
dy/dx
Std.Err.
t
p
dy/dx
Std.Err.
t
p
0.032
0.054
0.60
0.551
0.151
0.033
4.56
0.000
0.446
0.088
5.09
0.000
0.366
0.0465
7.87
0.000
0.219
0.049
4.47
0.000
0.246
0.159
1.55
0.121
0.446
3.582
0.12
0.901
19.261
5.210
3.70
0.000
-9.874
3.944
-2.50
0.012
0.043
0.078
0.55
0.582
0.033
0.037
0.91
0.363
0.369
0.282
1.31
0.190
0.062
0.040
1.57
0.117
0.115
0.052
2.23
0.026
0.059
0.179
0.33
0.742
-0.123
1.707
-0.07
0.943
-0.168
0.116
-1.45
0.146
-0.025
0.071
-0.35
0.723
0.065
0.184
0.35
0.724
0.535
0.097
5.52
0.000
0.283
0.111
2.54
0.011
0.981
0.310
3.17
0.002
3.471
10.006
0.35
0.729
-0.046
0.123
-0.38
0.706
0.276
0.065
4.22
0.000
0.764
0.172
4.44
0.000
0.470
0.088
5.32
0.000
0.208
0.087
2.39
0.017
-0.621
0.448
-1.39
0.165
0.524
7.373
0.07
0.943
sAcceptL*sCardDealL(%)
(1-sAcceptL)*sNoDealL(%)
sBeliefB(%)
sBeliefS(%)
market
location (SFU=1;UCI=0)
T16
T35
No. of obs.
1596
***
***
***
***
**
***
**
***
***
***
***
**
***
**
532
532
532
Notes: (1) dy/dx represents the marginal effect of the independent variable on the probability (in %) of a seller
accepting payment 2 (2) *p-value<=0.1; **p-value<=0.05; *** p-value<=0.01
Appendix: Experimental Instructions (T=2.8 treatment only; other treatments are similar)
Welcome to this experiment in economic-decision making. Please read these instructions carefully
as they explain how you earn money from the decisions that you make. You are guaranteed $7 for
showing up and completing the study. Additional earnings depend on your decisions and on the
decisions of other participants as explained below. You will be earning experimental money (EM).
At the end of the experiment, you will be paid in dollars at the exchange rate of 1 EM = $0.15.
There are 14 participants in today’s experiment: 7 will be randomly assigned the role of buyers
and 7 the role of sellers. You will learn your role at the start of the experiment, and remain in the
same role for the duration of the experiment. Buyers and sellers will interact in 20 “markets” to
trade goods for payment. There are two payment methods, payment 1 and payment 2.
Each market consists of two stages. The first is the payment choice stage. Each buyer is endowed
with 7 EM and decides how to allocate it between the two payment methods. Each seller is
endowed with 7 units of goods. Sellers have to accept payment 1, but can decide whether or not to
accept payment 2. Sellers who decide to accept payment 2 have to pay a one-time fee of 2.8 EM.
No participant observes any seller’s choice at this stage.
The second stage is the trading stage, which consists of a sequence of 7 rounds. In these 7 rounds,
you meet with each of the 7 participants who are in the opposite role to yourself sequentially and
in a random order. In each meeting you try to trade one unit of good for one unit of payment. The
buyer decides which payment to use and the trade is successful if and only if the seller accepts the
payment offered by the buyer. For each successful sale or purchase, you earn 1 EM less some
transaction costs. The transaction cost to both sides is 0.5 EM if payment 1 is used, and 0.1 EM if
payment 2 is used. If the buyer offers payment 1 (which is always accepted by sellers), then trade
is successful and both the buyer and the seller earn a net payoff of 1-0.5=0.5 EM. If the buyer
offers payment 2 and the seller has decided to accept payment 2 in the first stage, then trade is
again successful and both earn a net payoff of 1-0.1=0.9 EM. If the buyer has only payment 2 and
the seller has decided not to accept it, then no trade can take place and both earn 0 EM. At the end
of the market, unspent EMs or unsold goods have no redemption value and do not entitle you to
extra earnings.
Task summary
…
Stage 1: Payment choice
Buyers allocate 7 EM between the two payments
Sellers decide whether to accept payment 2 at a one-time fee of 2.8 EM
Stage 2: Trading (7 rounds)
Each buyer meets each of the 7 sellers in a random order
Trade with payment 1  net payoff of 0.5 EM
Trade with payment 2  net payoff of 0.9 EM
No trade  net payoff of 0 EM
Stage 1: Payment choice
Stage 2: Trading (7 rounds)
…
Market 20
Stage 1: Payment choice
Stage 2: Trading (7 rounds)
Market 1
Market 2
1
More Information for Sellers
As a seller, your earnings in a market (in EM) is calculated as
Option I
Accept payment 2
Option II
Not accept payment 2
Number of payment 1 transactions x 0.5
+ Number of payment 2 transactions x 0.9 – 2.8
Number of payment 1 transactions x 0.5
The benefit to sellers of accepting payment 2 is to increase the likelihood that you sell goods to
buyers (remember no trade can take place if the buyer has only payment 2 and you do not accept
it), and to reduce transaction costs and therefore increase net earnings by 0.4 EM each time a buyer
pays in payment 2. The cost to sellers of accepting payment 2 is that you have to pay a one-time
fee of 2.8 EM at the beginning of the market even if no buyers offer to pay you with payment 2 in
that market.
Which option leads to higher earnings depends on all other 13 subjects’ decisions. Table 1 on page
7 lists the average market earnings for the seller from the two options (accept / reject payment 2)
in cases where all buyers choose to allocate between 0~7 EM to payment 2, and where 0~6 of the
other 6 sellers choose to accept payment 2. As you can see, either option can give higher earnings
depending on other participants’ decisions. During the experiment, please keep Table 1 at hand for
reference. In addition, you can use a “what if” calculator on the computer screen to compute the
average earnings in situations where buyers make different payment allocations.
Your earnings from accepting payment 2 tend to increase if more buyers allocate more money to
payment 2, and if fewer sellers accept payment 2. The opposite is true if you reject payment 2.
More Information for Buyers
As a buyer, your earnings in a market are calculated as
Number of payment 1 transactions x 0.5 + Number of payment 2 transactions x 0.9
As a buyer, the benefit of allocating more money to payment 2 is that you save 0.4 EM each time
you use payment 2 instead of payment 1. The cost is the risk that you may not be able to trade if
the seller does not accept payment 2 and you run out of payment 1 (which is always accepted).
Your market earnings depend on your own payment allocation and the 7 sellers’ decisions on
acceptance of payment 2. Table 2 on page 7 lists the buyer’s market earnings if the buyer allocates
0~7 EM to payment 2 (and the rest to payment 1) and if 0~7 sellers accept payment 2. You should
allocate more money to payment 2 if you expect more sellers to accept it. Table 2 will also be on
your computer screen when you make payment decisions.
Forecast
At the start of each market before making payment decisions, you are asked to forecast other
participants’ choices for that market. Buyers forecast how many of the 7 sellers will choose to
accept payment 2. Sellers forecast (1) the average amount of EM that all 7 buyers will allocate to
payment 2, and (2) how many of the other 6 sellers will accept payment 2. You earn 0.5 EM per
correct forecast in addition to your earnings from buying/selling goods.
Earnings
At the end of the experiment, you will be paid your earnings in cash and in private. Your earnings
in dollars will be: Total earning (trading + forecasting) in EM x 0.15 + 7 (show-up fee).
2
Computer Interface
You will interact anonymously with other participants using the computer workstations. You will
see three types of screens (Figures 1-6 show sample screens).
Payment choice screen, Figures 1-2. This is where you make payment choices depending on
whether you are a buyer (Figure 1) or a seller (Figure 2). Each screen has 4 parts. The upper portion
summarizes information about previous markets. To the left of the blank column are your own
activities, including your payment choice, the number of transactions using each of the two
payment methods, the number of no-trade meetings, market earning from trading, and the number
of correct forecasts that you made. To the right of the blank column, there is an aggregate marketlevel statistic, the number of sellers who accepted payment 2.
The middle section provides information about your average potential earnings from trading in
each market. The buyer screen (Figure 1) shows Table 2. The seller screen (Figure 2) has a “what
if” calculator. A seller can type in the number of buyers choosing to allocate 0~7 EM to payment
2 and the number of other 6 sellers accepting payment 2 (the default value is 0 in all fields; the
first 8 fields must add up to 7; enter an integer 0~6 in the last field), press the “Calculate” button
to create a record showing the average market earnings from accepting payment 2 and not
accepting it, as well as the average buyers’ allocation to payment 2 in that scenario. For example,
if you would like to check your potential average earnings in the situation where 5 buyers allocate
2 EM to payment 2, 2 buyers allocate 3 EM, and 3 of the other six sellers accept payment 2, type
in “5” in the field “# buyers with pay2=2”, “2” in the field “# buyers with pay2=3”, and “3” in the
field “# other sellers accept pay2.” You can create as many records as you wish at the start of each
market.
In the lower-left section, you forecast what other participants will do in the new market. Enter an
integer within the indicated range for each forecast. The seller’s forecast of buyer’s average
payment 2 allocation is counted as correct if it lies within ±1 of the realized value.
In the lower-right section, you choose how to split your 7 EM between the two payment methods
if you are a buyer (Figure 1), and whether to accept payment 2 at a one-time fee of 2.8 EM if you
are a seller (Figure 2).
Trading screen, Figures 3-4. In each of the 7 trading rounds, buyers decide whether to buy a unit
of the seller’s good using either payment 1 or payment 2. This decision depends on the buyer’s
remaining balances of payment 1 and payment 2, and whether or not the seller has agreed to accept
payment 2; this information is shown on the buyer’s computer screen (see the lower left box in
Figure 3). Sellers do not choose at this stage, and can click on the “OK” button to review
information on the waiting screen (see Figure 4). From round 2 on, the upper section of the screen
reviews your activities in the previous round and in the current market up until then.
Waiting screen, Figures 5-6. At any point in the experiment if you finish your decision sooner
than other participants, you will see a waiting screen with information on previous markets and
your potential market earnings similar to what you observe on the payment choice screen.
Finally, sellers who invest in the one-time fixed cost to accept payment 2 may have a negative
“market earnings” in one or a few rounds. As a result of this, you may see a message screen
explaining the situation. After you have been alerted to this situation, you can click on the
“continue” button on the screen to proceed.
3
Figure 1: buyer’s payment allocation screen
Figure 2: seller’s payment 2 acceptance screen
4
Figure 3: buyer’s trading screen
Figure 4: seller’s trading screen
5
Figure 5: buyer’s waiting/information screen
Figure 6: seller’s waiting/information screen
6
Table 1: Seller’s average market earnings
•
•
•
This table considers the case where all buyers choose the same payment allocation;
use the “what-if” calculator for cases where buyers make different allocations.
The earnings for accepting payment 2 are in the upper-left corner,
The earnings for not accepting payment 2 are in the lower-right corner.
# of other 6 sellers accepting payment 2
All buyer’s allocation to
payment 2
0
1
0.7
0
0.7
3.5
2.1
3.0
2
2.5
3
3.5
2.0
3.5
1.5
5
1.0
1.2
1.4
3.5
0
3.5
0
2.3
0
1
2
3
4
5
6
7
3.5
1.2
3.5
3.1
1.8
3.5
3.5
0
0
1
2
3
4
5
6
7
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
3.5
3.9
3.4
2.9
2.4
1.9
1.4
0.9
3.5
3.9
4.3
3.8
3.3
2.8
2.3
1.8
3.5
3.9
4.3
4.7
4.2
3.7
3.2
2.7
3.5
3.9
4.3
4.7
5.1
4.6
4.1
3.6
3.5
3.9
4.3
4.7
5.1
5.5
5.0
4.5
3.5
3.9
4.3
4.7
5.1
5.5
5.9
5.4
3.5
3.9
4.3
4.7
5.1
5.5
5.9
6.3
3.5
3.5
0
# of sellers accepting payment 2
7
3.5
2.7
3.5
Table 2: Buyer’s market earning
Your allocation to
payment 2
2.3
3.0
0
3.5
3.5
3.5
0.9
3.5
0
2.6
3.5
1.8
3.5
1.9
3.5
3.5
3.5
0.7
3.5
2.9
3.5
3.5
1.5
2.1
3.5
2.6
3.5
0.6
3.5
3.5
2.1
3.5
3.5
←if accept pay 2
3.5 ←if not accept pay 2
1.1
1.6
2.4
3.5
3.5
3.5
0.5
7
2.8
1.2
3.5
3.5
0.7
3.5
1.8
2.8
3.5
3.5
3.5
6
3.5
1.8
3.5
3.5
6
0.7
3.5
1.3
2.1
3.5
2.3
3.5
4
3.5
2.9
3.5
1.4
2.6
5
0.7
3.5
1.6
3.5
4
0.7
3.5
3.5
3.5
3
0.7
3.5
3.5
1
2
0